I have an interesting finding in calculating the PSRR of this simple bandgap reference.
I would like to define PSRR=Vgs_p1/Vpower
If we have an ideal opam and these is no gain from Vcc to poam's output, we will have the PSRR as 1/loopgain. When the loopgain is low, the PSRR is bad.
However, if the opam is not ideal. Its output probably offset some PSRR and increase the supply rejection ratio.
Actually the PSRR you get here is very small. If you take the decibel value, it is negative. So, the more the gain of the opamp, more should be the PSRR because in your loop gain,which is Aβ, increase in A will decrease the PSRR which is negative becomes more negative.In fact, the PSRR is increasing here.
Actually, the way you would define PSRR for this is
PSRR = deltaVref/deltaSupply.
As far as P1 is concerned, You actually want the gate of P1 to move along with
the source.,which is the supply. So you can actually put a big cap from the gate
of P1 to the supply if you need a good AC PSRR.
If you are interested in PSRR in a DC sense, then you need to increase the lengths of all the top PMOS transistors.,(better cascode them if you have enough headroom)